Gradient-assisted radial basis function networks: Theory and applications

Kampolis, IC; Karangelos, EI; Giannakoglou, KC

dc.contributor.author	Kampolis, IC	en
dc.contributor.author	Karangelos, EI	en
dc.contributor.author	Giannakoglou, KC	en
dc.date.accessioned	2014-03-01T01:20:34Z
dc.date.available	2014-03-01T01:20:34Z
dc.date.issued	2004	en
dc.identifier.issn	0307-904X	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/15968
dc.subject	Multivariate interpolation methods	en
dc.subject	Noisy environment	en
dc.subject	Radial basis function networks	en
dc.subject.classification	Engineering, Multidisciplinary	en
dc.subject.classification	Mathematics, Interdisciplinary Applications	en
dc.subject.classification	Mechanics	en
dc.subject.other	Aerodynamics	en
dc.subject.other	Airfoils	en
dc.subject.other	Approximation theory	en
dc.subject.other	Combined cycle power plants	en
dc.subject.other	Functions	en
dc.subject.other	Gas turbine power plants	en
dc.subject.other	Interpolation	en
dc.subject.other	Mathematical models	en
dc.subject.other	Modal analysis	en
dc.subject.other	Thermodynamics	en
dc.subject.other	Activation functions	en
dc.subject.other	Radial basis function networks	en
dc.subject.other	mathematical method	en
dc.subject.other	mechanical engineering	en
dc.title	Gradient-assisted radial basis function networks: Theory and applications	en
heal.type	journalArticle	en
heal.identifier.primary	10.1016/j.apm.2003.08.002	en
heal.identifier.secondary	http://dx.doi.org/10.1016/j.apm.2003.08.002	en
heal.language	English	en
heal.publicationDate	2004	en
heal.abstract	This paper introduces a new variant of the radial basis function networks (RBFNs) with enhanced capacity to approximate any input-output mapping defined by a collection of activation signals and the corresponding responses. The new multivariate interpolation tool is conceptually drawn upon the standard formulation of RBFNs. However, the nonlinear mapping from the input to the hidden network units is modified by taking into account approximate values of the directional slopes of the response surface with respect to the free parameters. The RBF centers are selected in a forward manner and the activation function acts upon ""distances"" between input patterns and the RBF centers with components scaled by the aforementioned slopes. The latter should be viewed as local sensitivity measures simultaneously computed by the network itself. The improved performance of the new gradient-assisted radial basis function networks (GARBFNs) as interpolation tools will be demonstrated using a multimodal analytical function and two industrial applications related to the aerodynamic performance of airfoils and the thermodynamic performance of a gas turbine combined-cycle power plant. © 2003 Elsevier Inc. All rights reserved.	en
heal.publisher	ELSEVIER SCIENCE INC	en
heal.journalName	Applied Mathematical Modelling	en
dc.identifier.doi	10.1016/j.apm.2003.08.002	en
dc.identifier.isi	ISI:000187620700005	en
dc.identifier.volume	28	en
dc.identifier.issue	2	en
dc.identifier.spage	197	en
dc.identifier.epage	209	en